Oracle loses half its value while betting $95B on AI

Signals Inbox·July 15, 2026·AI Infrastructure

Oracle is preparing to spend roughly $95 billion on infrastructure in fiscal 2027, almost three times its previous annual capital spending, just as its shares have fallen about 47% since June 1. The tension is not whether customers want AI compute. Oracle has a record $638 billion backlog. It is whether one company can finance that demand without wrecking its cash flow and credit rating.

Q1What actually happened?

Oracle said in its official fiscal 2026 results that it expects to raise around $40 billion through debt and equity in fiscal 2027. S&P then cut Oracle’s rating from BBB to BBB-, the final step above junk. Meanwhile, Oracle’s shares fell roughly 47% from their June 1 high.

Q2Why is Oracle spending $95 billion?

Oracle needs data centers, GPUs, networking, power, and cooling to fulfill enormous AI cloud contracts. Its remaining performance obligations, basically contracted future revenue, jumped from $138 billion in June 2025 to $638 billion one year later. Oracle has found more demand than its current infrastructure can handle, so it is racing to build the capacity first.

Q3How unusually large is the spending?

Very large. Oracle spent around $35 billion in fiscal 2026 and could spend about $95 billion in fiscal 2027. That is almost a threefold jump in one year and more than Oracle’s entire $67.4 billion of fiscal 2026 revenue. Amazon, Meta, Microsoft, and Google also spend heavily on AI, but they enter the race with much larger cash flows and stronger balance sheets.

Q4Why did S&P downgrade Oracle?

S&P said Oracle’s AI infrastructure business brings rising capital costs, uncertain profits, fast-changing competition, and heavy customer concentration. Reports based on the rating decision say OpenAI represents roughly half of Oracle’s $638 billion backlog. If one giant customer delays projects or cannot pay, Oracle could still be stuck with debt, leases, and unfinished capacity.

Q5Is Oracle’s cloud business actually struggling?

No. That is what makes the story interesting. Oracle’s cloud infrastructure revenue grew 77% in fiscal 2026, reaching $18.1 billion. Its quarterly infrastructure growth reached 93%. Demand is accelerating. Investors are not rejecting the product. They are questioning how much debt, dilution, and negative cash flow Oracle must accept before that growth becomes profitable.

Q6How is Oracle different from other cloud giants?

Amazon, Microsoft, and Google built huge cloud businesses before the generative AI boom. Those businesses already generate cash that can fund new data centers. Oracle is trying to close the infrastructure gap much faster, partly through outside financing. It therefore has more concentrated customers, less room for delays, and a greater need for every new campus to fill quickly.

Q7So what is the real signal?

AI infrastructure is moving from a demand story to a financing test. Oracle has the contracts, rapid cloud growth, and one of the largest backlogs in tech. But it must spend almost $100 billion before much of that backlog becomes revenue. If Oracle succeeds, it becomes a serious fourth hyperscaler. If it fails, this could become the clearest warning that AI compute promises grew faster than the balance sheets supporting them.